On behalf of the Information Theory Group at Stanford University, we welcome you to learn about our latest developments in 2009! Here you will find a set of interactive applets that will perform lossy compression before your eyes! No download is necessary, and it is free of charge. You are in control of key parameters through these dynamic pages, and you can view the outcome at the click of a button!
We have developed lossy compression through two approaches: Markov Chain Monte Carlo and Viterbi Algorithm. To begin your journey, click on a tab 'MCMC' or 'Viterbi', to find these interactive applets.
A nonconvex optimization solution via Simulated Annealing
Lossy compression via Viterbi encoding
Although lossless compression may have been perused, very few state of the art technologies exist for lossy compression. Huffman coding and Run Length Encoding are the two primary, if not only, lossy compression algorithms for a binary sequence.
The Mars Exploration Rovers use ICER, a discrete wavelet transform based image compressor. But even at the heart of a state of the art lossy compressor, after a set of transform coefficients is calculated, the final binary sequence is still lossless compressed. The reason for this is that no good lossy compressor exists for a binary sequence; we are proposing there exists one, and we are displaying our work as a first step in convincing you.
Disclaimer: the applets can be used for academic or educational purposes but should not be abused for commercial usages where a financial stimulus is the primary motivation.